DE69023787T2 - Target location on a target. - Google Patents

Description

The invention relates to a method for locating a target point on a target object and in particular, but not exclusively, to a skeletal target point localization which can be carried out in parallel processing devices.

GB-A-2 202 112 relates to a method for identifying regions of interest in a binary image of a viewed scene. US-A-4 796 187 describes a method for selecting an aiming point on a target viewed by an imaging system. The method involves examining various images in a viewed scene and selecting those elements which are permanent and near the center of the scene for further processing. DE-A-2 949 453 describes a method for improving missile accuracy by comparing a target image with a stored image in a correlation circuit.

Applicant has developed a target tracking system in which a digitized image of a target and its background are generated, and the image is then processed to obtain a skeleton image of the target. The term skeleton image used herein is intended to indicate an image consisting of one or more linear elements which form a kind of 'pin figure' representation of the original image. In an embodiment to be described, the 'pin figure' comprises horizontal and vertical linear elements one pixel wide, each corresponding to features, for example a central region and lobes of the target image. At least approximately, the linear elements are symmetrical. centered between boundaries of the target on the two sides of the element. In this embodiment, the skeleton essentially comprises all the information contained in the original image, ie the original image could be reproduced by a skeleton. This follows from the algorithm used to form the skeleton.

It may be desirable to assign a specific pixel within the target image, namely the center of mass or the centroid of the image, to an aiming point for a weapon guided to the target.

An object of the present invention is to provide a quick and simple method for determining the aiming point of a target using parallel processing devices.

According to a feature of the invention, a method for determining a target point in a projectile guidance system comprises the following steps:

an image of the scene being viewed is created;

determining the presence of a target within the scene and generating a digitized binary image of the scene, distinguishing the target from the background of the scene;

the method being characterized by the following further steps:

contour mapping the binary image to create a contour image of the target and creating a separate skeleton image of the target using a recursive process, the skeleton having substantially the thickness of one element; the skeleton image is contoured to create a contoured skeleton; and

Parallel processing devices are used to create a centroid from the contour values and the position of the skeleton elements and to determine the element closest to the centroid, with the element closest to the centroid forming the target point of the target object.

An embodiment of the invention will now be described with reference to the drawing. In the drawing:

Fig. 1 illustrates a binary image of a target;

Fig. 2 shows how the image according to Fig. 1 is designed in terms of the outline or its contours;

Fig. 3 illustrates a finished outline image corresponding to Fig. 1;

Fig. 4 shows an image equivalent to Fig. 3 but satisfying the skeleton image conditions;

Fig. 5 illustrates the skeleton image according to the target of Fig. 1; and

Fig. 6 shows part of another skeleton image, where the contour values are assigned to the pixels of the skeleton.

To determine the aiming point at a given target, a skeleton image was created from the image of a viewed scene containing the target. A binary image of the target is created by any conventional method known to those skilled in the art. The binary image is shown in Fig. 1 and comprises an array of pixels, each of which has a value of either '0' (corresponding to the background) or '1' (corresponding to the target).

The image according to Fig. 1 is provided with a contour, i.e. an outline, to create a value for each pixel of the target area. To do this, the field is scanned in a raster manner until the first '1' is located. This '1' is the starting point for the first contour, namely the boundary contour. By moving clockwise, all the fixels on the boundary are assigned the contour value '2' (Fig. 2). The field is again scanned as before until the first '1' is located, and then all the pixels as before that have the value '1' within the boundary contour (with the value '2') are assigned the value '4' if the movement is in a clockwise direction. This process is repeated, assigning each of the inner contours a value that is two greater than its adjacent outer contour, and this continues until no more pixels with the value '1' are located. The finished contour image is shown in Fig. 3.

At the same time as the contour image is being generated, another 'image' of the binary field is generated. Similar processing is used to that used to produce the contour image, but different 'test' conditions must be met for each pixel. As described above, the first '1' is located by raster-scanning the binary field. When a '1' is located, a 3 x 3 operator is added, and the operator is:

Pixel C is the one to which the 'test' conditions are imposed, and in this case it is the first '1'. The other pixels of the operator each take the values of the eight adjacent pixels, unless any of those pixels does not have the value '1'. For each such pixel that does not have the value '1', the relevant pixel of the operator is given the value '0'.

For the first '1' we get:

Then the 'test' conditions are applied.

=> (i) more than two of the possible (A + B) pixel pairs must be equal to 1, i.e. A + B = 1 for at least three pairs of adjacent pixels;

=> (ii) the sum of all A's and B's must be less than three.

If either or both conditions are met, the pixels located at C remain unchanged. If neither condition is met, then the value of the fixel at location c is given by an augmented value in the same way as the values were assigned to the pixels when the contour image was created as described above. In the above example, neither condition is met. Therefore, the pixel takes a contour value of '2'. If later, during the contouring, another pixel with the value of '1' is taken:

All values of A and B are not '1', so the operator elements corresponding to the non-'1' image elements are made zero:

In this case, condition (ii) is satisfied and therefore pixel C (in the center) remains unchanged, i.e. the value of '1' is maintained. After this process is completed for each pixel, the array shown in Fig. 4 is obtained, i.e. a contour image except where the 'test' conditions are satisfied. From this array, the skeleton image is generated where all pixels not having the value "1" are assigned the value of '0' and this array is shown in Fig. 5.

The field of Fig. 5 can be contoured by any suitable method to produce a field corresponding, for example, to the field of Fig. 6. The skeleton pixels can simply be assigned to the contour values that have already been determined for the corresponding pixels of the previously generated contour image.

According to Fig. 6, to determine a suitable target point location or the centroid of the target, an xy axis system is centered on one of the pixels, preferably one with the highest skeleton contour value, i.e. the skeleton centroid. The coordinates and of the actual centroid are then determined from the following equations:

= ΣΣ x C(x,y)/ΣΣC(x,y)

and

= ΣΣ y C(x,y)/ΣΣC(x,y)

where x and y are values of the axis system and C(x,y) are the contour values found at the skeleton pixel locations. From and the distance D of the actual area centroid (marked with *) from the skeleton area centroid can be determined, i.e.:

D = ( ² + ²)

Now consider the skeleton shown in Fig. 6. The centroid C is marked and has the contour value of 14. The x,y axis is centered on this point. ΣΣC(x,y), ΣΣxC(x,y) and ΣΣyC(x,y) are determined as shown.

ΣΣC(x,y) = 6+8+10+12+14+12+10+8+6+4

=90

Σ Σ C(x,y) = (-4x6)+(-3x8)+(-2x10)+(1x12)+(0x14)+(1x12)+

(2x10)+(3x8)+(4x6)+4x4)

= -24+ -24 -20 -12 +12 +20 +24 +24 +16

= 16

ΣΣyC(x,y) = (0x6)+(0x8)+(Ox10)+(0x12)+(0x14)+(0x12)+

(0x10)+(0x8)+(1x6)+(2x4)

= 6 + 8

= 14

= ΣΣxC(x,y)/ΣΣC(x,y)

= = 0.1778

90

= ΣyC(x,y)/ΣΣC(x,y)

= = 0.1555

90

The distance D from the actual centroid of the skeleton surface is also determined:

D = ( ² + ²)

D = (0.1778² + 0.1555²)

D = (0.0316 + 0.0242)

D = 0.236

If necessary, the distance D for each skeleton pixel can be determined so that the pixel closest to the centroid can be determined, and this pixel is used as an indicator of the target point location.

Claims (8)

1. Method for determining the aiming point of a target in a missile guidance system, which comprises the following steps: an image of a viewed scene is created; detecting the presence of a target within the scene and generating a digitized binary image of the scene, with the target distinguished from the background of the scene; characterized by the following further steps: contouring the binary image to create a contour image of the target, and generating a separate skeleton image of the target using a recursive process, the skeleton having a thickness of substantially one element; the skeleton image is captured in contour form to create a skeletal contour; and a parallel processor is used to determine a centroid from the contour values and the position of the skeleton elements and to find the element that is closest to the centroid, whereby the element closest to the centroid represents the target point of the target. 2. The method of claim 1, wherein contouring the binary image includes the step of assigning to each element of the image a value indicative of the location of the element. 3. Method according to claims 1 or 2, in which the recursive method comprises the following steps: the image is scanned in a raster fashion using a 3 x 3 operator; the elements of the image are assigned values that correspond to the central position of the operator, and the assignment is made by the position of the element within the image; and the value zero is assigned to all those elements that were not assigned a '1' in the previous step, thus creating a binary skeleton image. 4. A method according to claim 1, wherein the binary image is contour processed by assigning to each element of the target within the binary image a value corresponding to the distance of the element from the perimeter of the target image. 5. A method according to claim 1, wherein the skeleton image is generated by assigning to each element of the target within the binary image a value corresponding to the distance of the element from the perimeter of the target image and a function of the contour value of each element adjacent to that pixel if that pixel has been assigned a contour value or a binary value corresponding to the binary image of each element adjacent to the pixel if the pixel has not been assigned a contour value. 6. Method according to one of the preceding claims, in which the contour-processed skeleton image is generated from a combination of a contour image and a skeleton image. 7. Method according to one of the preceding claims, in which the coordinates of the area center of gravity relative to an axis system centered on an element of the contour skeleton are each calculated by corresponding mathematical functions which have a plurality of expressions, one of which is composed of the respective contour values of the contour skeleton image and another expression indicates the distance of the respective elements according to the axis system from that element on which the axis system is centered. 8. Method according to one of the preceding claims, in which the location of the centroid is used to determine the element of the skeleton which is closest to the centroid.